抗几何攻击的数字图像水印研究

【作者】 付正立；

【导师】 李建平；

【作者基本信息】 电子科技大学 ， 信息与通信工程， 2009， 硕士

【摘要】 随着Internet技术和计算机技术的应用和迅猛发展,人们在日常生活和工作中越来越离不开数字技术和网络技术。数字化和网络化在给人们带来便捷和高效的同时,也引发了一些新的问题等待科学工作者去研究和解决。其中数字版权保护问题就是由此引发而来的一个棘手和富有挑战性的课题。数字版权保护除了使用常规的密码技术做加密处理之外,在很多领域和场合更需要一种称为数字水印的技术来发挥更广泛和有效的作用。而大多数数字水印算法在抗几何攻击方面都不具备鲁棒性,本文工作基于此而展开。本文主要介绍了数字图像水印技术的起源、发展历史、现状、常用的模型、分类方法、衡量标准。实现了一个结合SVD和DCT的能有效抵抗几何攻击(剪切、旋转、镜像、缩放、压缩等)的数字图像水印算法。本文重点研究基于变换域的抗几何攻击的数字图像水印算法,主要的工作有:(1)研究了离散余弦变换域内的常用水印嵌入算法,通过对其进行常见的几何变换攻击测试,找到了DCT域内的数字图像水印算法抗几何攻击鲁棒性差的根本原因是水印信息同步性差。在此基础上,研究了一种DCT域的第二代水印算法,该算法采用特征块来嵌入水印信息,较好的解决了水印信息同步的问题。在大量实验的基础上,对该特征块选择算法进行了改进;实验表明,改进后的算法提取到的特征块更利于嵌入水印信息。(2)重点研究了奇异值分解方法在抗几何攻击方面的特性和应用。通过实验,验证了基于块嵌入的SVD域内水印算法能有效的抵抗常见的几何攻击手段,并指出了其在抵抗滤波攻击等方面存在的不足。然后在此基础上设计并实现了一种结合DCT和SVD的数字图像水印算法,该算法结合了DCT的不可见性和SVD的抗扰动性。经过测试,该算法能有效的抵抗常见的几何攻击,并且具有较高的水印嵌入容量。(3)为了增强在抗几何攻击方面的精确性,研究了基于几何矩的图像几何变换特征参数提取算法。该算法计算量小,精度比较高,可以用于水印检测算法中对待检测图像的预处理。实验表明,经过有效的预处理,能提高水印检测的精度。更多还原

【Abstract】 With the development and application of Internet and computer technology, digital media and network communication have become to a basic facility in our daily life and work. These technologies make our life more convenience and efficiency. But these technologies also bring some new problem in another side, such as digital copyright protection. Digital copyright protection is a challenging task, because the cryptology technology can only do protect the digital media on transform. The digital watermarking technology can protect the media even if the media has been decrypted. Almost of the watermarking method do not have robustness on geometric-attack, this thesis do some research work on resistant geometric-attack watermarking.An introduction of the digital watermarking’s origins, development history, current research status, communication models, classical algorithms, classification, and benchmark have been given. A new watermarking algorithm based on DCT and SVD have been designed and implemented, this algorithm has geometric-attack resistant and invisible-watermark advantages.This thesis focus on the anti-geometric-attack watermarking in transform domain, the main research works are below:(1) A research on the DCT domain watermarking algorithm. Most DCT domain watermarking algorithm can not have robustness on geometric-attack, and the root cause is: after some geometric-transform, the watermark synchronization is not work. A analysis of the second generation watermarking algorithm on DCT domain have been done, which uses the character block to embed the watermark information. At last an improvement of the block selection algorithm has been done, to make the insert-method more efficiently.(2) A research on the SVD transform domain watermarking, which has strong robustness on geometric-attack. Some experiments and verifications show this feather. A block-based watermarking system with the combination of SVD and DCT has been done. This system can combine the non-visibility of DCT and the anti-disturbance of SVD.(3) An enhancement of the accuracy of anti-geometric-attacks. The correction method of geometric-transform parameters was introduced. The method is based on the 0 and 1 level geometry moment, and can compute quickly and correctly.更多还原